U.S. reportedly orders Anthropic to suspend model access, sparking criticism that the startup's safety lobbying backfired

No. The simple fact is that once you wake up Leviathan, it is not always going to behave the way you want it to. It can arbitrary and capricious.

Urging regulation *only in this specific way* was never a good strategy.

the gleeful “stupid games, stupid prizes” response to the Anthropic situation is wildly shortsighted.

you can dislike Anthropic’s choices and still recognize that opaque, sweeping government interventions against frontier models doesn’t bode well for any ai company.

The government should not be regulating AI to this extent. Not like this.

I’ve been against onerous regs when Anthropic and the safety community was pushing for it.

And I’m against it now that they got what they asked for.

If you keep turning up the temperature by claiming you're building a cyberweapon or a nuke, while racing towards RSI, and you are begging the government to stop you and slow things down, you should not be surprised when they eventually do so (and not in the way that you want!).

If you cannot be *extremely prescriptive* in what you want, generally pushing the government to do things closer to what you want ("regulate bad things") is quite unlikely to get you what you want. I feel this lesson is so obvious as to not need mentioning but clearly that's not true.

@nabeelqu the function of a sovereign is that you don't get to dictate terms with them, especially once you poke at them incessantly

I don’t think it’s too much to ask for Anthropic to stop fucking with its customers (major risk for AI users) AND for the state to stop fucking with Anthropic (major risk for corporate governance norms writ large)

The position "cannot be jailbroken at/above this capability level" (which is far from AGI btw) is not a stable policy to hold for many reasons. I believe the latest skirmish is net positive for Anthropic (and net negative for future innovation). USG/Anthropic may have a real disagreement but the USG is engaging on the battlefield right where Anthropic wants it (regulation).

@nabeelqu not urging any regulation, also, was never a good strategy

@levie Yes. It is a huge problem - the doomer takes are threatening us all.

Doomer fears never materialize historically; regulatory damages always do

Counter argument:

The message to the world yesterday was: "Be very careful before taking dependency on such models for anything business critical, because someone can pull the plug at any minute, without warning and for whatever reason". This translates to zero sovereignty.

Would another government agency take dependency on this? Absolutely not.

Would any major business take dependency on this? No way.

It is not a risk they can afford to take.

@nabeelqu Guilty until proven innocent is a rational approach for the gov to take if Anthropic is loudly implying this model is the start of the existential risk era.

@NatPurser Yeah like knowing Trump’s intentions to burn down everything to the ground this isn’t something that really benefits anyone

@BogdanIonutCir2 @nabeelqu god forbid we ask for measured, targeted, smart regulation. the audacity

@nabeelqu seems strange to put the blame on Anthropic and not on the unreasonableness of the admin, though, in this case

@krishnanrohit @nabeelqu or, to be more precise, I don't think setting a standard where AGI labs are expected to lie for consequentialist reasoning related to what the USG would (unreasonably) do otherwise would be a good development, especially with A[G]I

@BogdanIonutCir2 I am also confused by your first clause, they have been pretty clear about the risks from Mythos and justified a closed rollout on the basis of those risks

@BogdanIonutCir2 Indeed, it is almost as though there was a virtuous middle path here!

@nabeelqu can you point to some middle path you think would be better than what Anthropic was asking for?

@drydenwtbrown I mean see their blog post or just talk to your friends at the company?